AI is ‘Accelerating the Climate Crisis,’ Expert Warns

AFP/APP

Montreal: If you care about the environment, you might want to reconsider how you use artificial intelligence (AI). Generative artificial intelligence consumes 30 times more energy than traditional search engines, cautions researcher Sasha Luccioni, who is dedicated to raising awareness about the environmental impact of this burgeoning technology.

 Recognized by Time magazine in 2024 as one of the 100 most influential figures in AI, the Canadian computer scientist has spent years quantifying the emissions of programs like ChatGPT and Midjourney.

“I find it particularly disappointing that generative AI is being used just to search the Internet,” said Luccioni, speaking at the ALL IN AI conference in Montreal. She explained that the language models underlying these programs require vast computational power to train on billions of data points, demanding energy-intensive servers. Additionally, significant energy is expended in responding to each user query.

Unlike traditional search engines that merely retrieve information, AI models create new information, making the process “much more energy-intensive,” Luccioni explains.

According to the International Energy Agency (IEA), the combined energy consumption of the AI and cryptocurrency sectors reached nearly 460 terawatt-hours of electricity in 2022 — roughly 2 percent of global production.

Energy Efficiency Efforts

A leading expert on AI’s environmental impact, Luccioni co-developed the CodeCarbon tool in 2020 to help developers quantify the carbon footprint of their code. This tool has since been downloaded over a million times.

Now the head of climate strategy at Hugging Face, a platform for open-access AI models, she is developing a certification system to measure algorithms’ energy efficiency.

This system would function similarly to the US Environmental Protection Agency’s energy ratings for devices, allowing users to assess the energy consumption of AI models. “We don’t factor in water or rare materials,” Luccioni admits, “but we can measure energy efficiency and rate models, giving one an A+ and another a D.”

A Call for Transparency

Luccioni has been applying her tool to open-source generative AI models, but she is eager to evaluate commercial models from companies like Google and OpenAI, who have been less forthcoming.

Despite pledges from Microsoft and Google to achieve carbon neutrality by 2030, the greenhouse gas emissions from these tech giants surged in 2023, driven by AI expansion—up 48 percent for Google since 2019 and 29 percent for Microsoft since 2020.

“We are accelerating the climate crisis,” Luccioni warns, calling for greater transparency from technology companies. She argues that governments are currently “flying blind,” with little insight into the data sets or training processes behind AI models. Transparency, she believes, is essential for enacting meaningful regulation.

Advocating ‘Energy Sobriety’

Luccioni also stresses the need to educate the public about what AI can and cannot do, and at what environmental cost. In her latest research, she found that generating a high-definition AI image consumes as much energy as recharging a smartphone battery.

As AI becomes more integrated into everyday technologies—from conversational bots to connected devices—Luccioni advocates for “energy sobriety.” The goal, she emphasizes, is not to reject AI outright, but to select and use AI tools thoughtfully.